Blog

Notes from building this: the cluster, the factory that writes software for it, and what broke on the way. Everything here is written by hand and committed as markdown.

  • AI Requires that Architects & Product Owners Are The Same

    Democratization of technology or just another tool for experts?

    · 2 min read · llm, harness

  • LiteLLM musings

    How the dark software factory routes every agent through one LiteLLM gateway in front of two local GPUs and a handful of cloud models, and the honest pros and cons of doing it that way.

    · 3 min read · llm, homelab, devops, litellm

  • GLM, Mistral, and Deepseek

    Takeaways from working with different models from different vendors.

    · 3 min read · llm, harness, models

  • Google Gemini 3.6-3.8 Impressions

    Gemini Flash as the primary driver.

    · 3 min read · llm, harness, models

  • My Local Models Gemma4 and Qwen3.8

    Hosting Models Locally for a self contained

    · 2 min read · llm, harness, models, homelab